一种基于自编码神经网络的快速行人检测算法
A Novel Fast Pedestrian detection Algorithm based on Auto-encoder Neural Network
投稿时间:2017-10-29  修订日期:2017-10-29
DOI:
中文关键词: 行人检测  方向梯度直方图特征  AdaBoost算法  自编码网络  ACF模型
英文关键词: Pedestrian detection  HOG feature  AdaBoost algorithm  Auto-encoder network  ACF model
基金项目:河北省高等学校科学技术研究项目(ZD2016158); 国家青年科学基金项目(F2015204130); 省科技项目基金(14227404D)
作者单位E-mail
李得峰 河北农业大学 ldfeng84@hotmail.com 
韩宪忠 河北农业大学  
摘要点击次数: 6
全文下载次数: 
中文摘要:
      针对传统基于方向梯度直方图(HOG)特征与AdaBoost算法分类器在目标检测中存在检测速度慢、误差率大的问题,本文提出了一种鲁棒新颖的快速行人检测算法。该算法首先利用基于ACF(Aggregate Channel Features)模型的目标检测算法对图像进行预处理,获得疑似目标区域;然后对获取的子区域进行尺度归一化,提取HOG特征,并输入到自编码神经网络中进行降维;最后利用Adaboost分类器对分类检测,输出检测到的行人区域。大量的定性定量实验表明本文所提算法的行人检测性能超过现有的检测算法,其检测速度也超过大多数算法。
英文摘要:
      Since the traditional algorithm has the advantages of slow detection rate and large error rate in pedestrian detection, a fast pedestrian detection algorithm based on auto-encoder neural network and Adaboost is proposed in this paper. Firstly, the pedestrian detection algorithm based on ACF model is used to process the image so as to obtain the Suspected object area. Then, the acquired sub-region is normalized and the HOG feature is extracted, input into the auto-encoder neural network. Finally, the Adaboost classifier is used to detect the classification and output the detected pedestrian area. A large number of qualitative and quantitative experiments show that our proposed method has more performance than the existing detection algorithm for pedestrian detection, and its detection speed is also more than most of the algorithms,which is suitable for engineering application.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭